Assessing the validity of ’s 400- year gazetteer record for establishing historical baselines

Connor Walsh August 2016

A thesis submitted for the partial fulfilment of the requirements for the degree of Master of Science at Imperial College London

Submitted for the MSc in Conservation Science 2

Declaration of Own Work

I declare that this thesis: Assessing the validity of Hainan’s 400-year gazetteer record for establishing historical baselines is entirely my own work, and that where material could be construed as the work of others, it is fully cited and referenced, and/or with appropriate acknowledgement given.

Signed. Name of student: Connor Walsh Name of Supervisors: Dr Samuel Turvey; Clare Duncan

3

Table of Contents Declaration of Own Work ...... 2

List of acronyms ...... 4

Acknowledgements ...... 4

Introduction ...... 6

Methods ...... 8

Results ...... 12

Discussion ...... 16

Supporting Information description ...... 20

Literature cited ...... 21

Appendix S1. Chinese characters and transliteration ...... 24

Appendix S2. Digitization workflow ...... 26

Appendix S3. Raw data: gibbons ...... 27

Appendix S4. Raw data: other species ...... 85

4

List of acronyms CE: Current Era OLE : Optimal linear estimation

Assessing the validity of Hainan’s 400-year gazetteer record for establishing historical baselines Abstract Conservationists have sampled the enormous collection of local gazetteers in to find species information dating from as much as 1,500 years ago until the 20th century. However, the inherent statistical completeness and biases of the gazetteers, and their ability to provide a robust historical baseline of past biological conditions and faunal changes has yet to be tested. This paper collates gazetteer references to the world’s rarest ape, the Hainan gibbon, and other mammal species on Hainan island, China’s southernmost province. The records cover 1521—1935, with 114 gibbon records. The density of records increases through time towards the present. After qualitatively scoring these records they were analyzed with optimal linear estimation (OLE), a technique to estimate extinction date based on limited historical sightings. The findings show the gazetteer record is robust because it predicts survival for species that are still extant today, past the end of the gazetteer record. This suggests gazetteers may provide useful information on past environmental conditions and change. The gazetteers also record wolf and dhole, species that are not otherwise confirmed to have occurred on the island. Further OLE on those 2 species suggests they may have been present on Hainan well into the 20th century.

Keywords: Dhole, Hainan gibbon, optimal linear estimation, taxonomic boundaries Word count: 5967 5

Acknowledgements

The initial concept for this study, and much of the context and hope for a practical impact, comes from S. Turvey of the Institute of Zoology (IoZ). C. Duncan also of IoZ guided statistics. B. Fuehrer of SOAS, University of London, provided sinological and translation guidance. C. Searle of Renmin University of China helped source Chinese documents not available in Europe, including the base text. Introduction environmental destruction in the 20th and 21st centuries. In the modern era China has Historical baselines are used in seen repeated waves of environmental conservation to help make sense of current damage: since the founding of the People’s environmental patterns and processes, inform Republic of China in 1949 the human management decisions, and place population has doubled, post-war economic conservation into wider cultural and prioritization led to land conversion, which intellectual contexts (Szabó 2010; Rick & was followed by the Great Leap Forward Lockwood 2013; Davies 2014). Shifting from 1958 to 1961, another period of major baseline syndrome, where the perspective of environmental damage (Edmonds 1999); and what is ‘normal’ is degraded over time, can subsequent economic growth spurts in the mislead not just on population size (Pauly 1980s, 1990s, and 21st century were not 1995) but even the possible former regional matched by conservation action (Liu & existence of large mammal species. Diamond 2005). The Environmental However, historical sources seldom match Protection Law came into effect only in the requirements of a contemporary 1989. With the likely loss of the biodiversity survey (Rick & Lockwood river dolphin or baiji (Lipotes vexillifer), 2013): records may be incomplete, organisms China is the scene of one of only 2 may be described according to folk extinctions of an entire mammal family in classifications which ignore or pre-date the past 2 centuries (Turvey 2009). In this standard scientific nomenclature (Fleck et al context native biodiversity baselines are 1999); and over centuries the individual urgently needed. observer will change many times. Therefore while historical baselines are valuable, the Previous investigations into the past sources must be critically evaluated before status of Chinese biodiversity using historical they can confidently be applied to inform sources have been conducted by Wen (2009, conservation science and management. 2013), Chatterjee et al (2012), and Turvey et al (2015b). Their approaches began in the With its ancient written record scholarly tradition of sinology (the study of (Wilkinson 2015), China appears to hold pre-modern China) and advanced to strong potential as a place to identify conservation biology with GIS and statistics, historical baselines. The need is urgent: potentially identifying the most appropriate anthropogenic pressures have led to extreme habitat types to restore, and predicting 7

responses to climate change (Chatterjee et al In this study the biases and errors to 2012; Turvey et al 2015b). However before contend with include inexpert observations; historical baselines can be constructed using editorial characteristics of the gazetteer style; historical texts, the original records still need irregular publication dates; and different to be critically examined to see just how ideas of what constitutes a species. A robust a source they are. qualitative process involving historical, linguistic, statistical and taxonomic steps can Of all the biodiversity challenges in aspire to see past these biases. China, arguably the most urgent is that of the Hainan gibbon (Nomascus hainanus), the Identifying species in these historical rarest ape on Earth, with fewer than 30 sources is resource-intensive. While the individuals living, limited to just one nature gazetteers list a range of fauna and flora, I reserve, on the island of Hainan (Turvey et al have restricted this study to mammals. The 2015a). In this situation any new information various authors’ understandings of plants and on the species’ ecology, distribution, and animals will differ greatly, both between dynamics are valuable. In this paper I kingdoms and temporally, therefore analyze historical gazetteer records from comparing records from different taxa over Hainan Province dating from 1521 to 1935 to time may introduce additional biases. Only assess mentions of Hainan gibbons and other comparing mammals aims to limit this. mammals over time, with the broader goal of examining the validity of the sources that The spatial focus of this study, previous studies have taken when using Hainan, is a tropical island province of about gazetteers to inform Hainan gibbon 34,000 km2. It lies south of the Chinese conservation (Chatterjee et al 2012; Turvey mainland at about 20º north and as close as et al 2015b). There are no known Hainan 240 km from Vietnam. The region it sits in, gibbon population estimates to compare with Southeast Asia, has (along with South Asia) before the late 20th century (Turvey et al the highest density of threatened land 2015a), so the analysis is based on whether mammals (Schipper et al 2008). In the 20th the animals were likely extant in 1935 (by century Hainan faced huge environmental virtue of being extant today). This can simply pressures, with 35% of potential gibbon indicate whether the gazetteer record is habitat lost between 1991 and 2008 (Zhang robust. et al 2010). Potentially confirming the gazetteers as a valid tool for researching 8

local historical baselines may open up to the taxable capacity of a county, biological and social conservation strategies, prefecture, or province, they include relevant meeting an urgent need. local biographies, culture, topography, and produce (Bol 2001; Wilkinson 2015). The To test the Hainan gazetteers I apply scattered books and fragments of Hainan’s optimal linear estimation (OLE) (Roberts & historical gazetteers were located, edited, and Solow 2003; Solow 2005) to the qualitatively published in 2003 as a collection of 68 scored sightings of gibbons and other hardback volumes (Hong 2003). After a mammals. Comparing the OLE extinction digitization process (see Supporting estimates, based on gazetteer data, to the Information for the full process) I had 39 otherwise known presence (or not) of species machine-readable volumes dating from 1521 can indicate the internal coherence of the to 1938 which form the reference text for this gazetteer sightings record. study. I discarded 23 corrupted volumes and 6 missing volumes as bad data. The final In this paper I evaluate the Hainan volume, from 1938, was also discarded gazetteers as a potential source for because it deals exclusively with other establishing species historical baselines up to islands under the administration of Hainan, 1935. To do this I synthesize and analyze but not Hainan island itself. This leaves 1935 data on past species records in the Hainan as the final date in the study for which gazetteers, to then evaluate the validity of potential environmental data are available this historical archive at predicting species from the gazetteers. The 38 volumes formed persistence. I then use that assessment to a corpus which I searched with AntConc assess records of otherwise unconfirmed version 3.4.4 (Anthony 2014). My initial mammals on Hainan. keyword searches were based on terms for Methods apes or gibbons, two variants of yuán (see Supporting Information for the Chinese The local gazetteer is a historical characters of the search terms). I collected genre in China of which 8,584 editions exist these entries in a spreadsheet along with the up to 1912, covering 1,500 areas of China gazetteer name, publication year, and some neighboring countries (Wilkinson administrative unit, date, specific 2015). They often served as hand-over geographical information, page number, documents for civil servants assigned from translation, and the other mammals listed in elsewhere in China. As a need-to-know guide the surrounding paragraphs (see Supporting 9

Information for a reduced version of the I scored the entries for reliability spreadsheet). I checked that places named in according to the following criteria: the records were indeed on Hainan—it was at EX: Record exists but it is excluded from times administered as a single unit with the analysis, because it describes an animal nearby mainland provinces, necessitating which is called an ape but is definitely not a confirmation of which entries were spatially gibbon (e.g. it has a tail; it is the wrong relevant. The spatial scale of the gazetteers is color); or is identifiably referring to another variously county, prefecture, whole-island, or part of China. province. The quality of maps is in the 0: no gibbon record of any kind available for gazetteers is also variable, with internal a given gazetteer. Counted towards gibbon boundaries not appearing until 1930 (Fig. 1). absence data. Thus the spatial unit in this study is the 1: Record is mythical or fantastical, e.g. island of Hainan, regardless of the individual inkpot gibbon, fantastical poetry, or gazetteer’s unit. extremely limited, such as place names incorporating ‘ape’. E.g.: In addition, Figure 1: Historical maps of Hainan created in 1689 Qiongzhou [Hainan] has stone apes, the (Hong 2003) and 2007 (CHGIS 2007) indicate the small ones the size of a fist, they grow when spatial difficulties inherent in using the gazetteers for fed water from a well. Also [known as] black identifying historical land mammal range.

apes, they can rub ink [from an inkstone], jumping into the ink pot when finished.

10

2: Record contains a brief mention of a I identified the book-sections which gibbon, but with no or very limited detail; usually had gibbon references, searched for record consists of ‘naturalistic poetry’; or sections with those titles across the corpus, record is a standard text quoted from an and identified those that did not mention earlier gazetteer. E.g.: Gibbons: there are gibbons. I then further recorded every three types: Golden ones are yellow, jade- mammal in the ‘beasts’ (shòu) sections faced are black, and there are also pure across the corpus. I identified all of the black ones. The golden and jade-faced ones mammal species possible by cross- are all rare. referencing historical Chinese dictionaries 3: Record is original, and contains relatively ( 1948; Guo 1981; Hanyu dacidian detailed description of either gibbon 1986; Guhanyu changyongzi zidian characteristics/biology and/or geographical bianxiezu 1998; Duan 1997), 19th century location of gibbons. E.g.: Gibbons: The male British accounts (Swinhoe 1870) and the is black and the female pale. They are like most comprehensive modern Chinese monkeys but their arms are very long, mammal guide (Smith & Xie 2008), and then climbing and grabbing branches in the conducted new full-corpus searches with canopy with great agility. If they ever fall to them. This was a more rapid system than that the ground they stiffen up like a tree. As they used with gibbons: I counted only entries are good at climbing trees but cannot walk, which scored equivalent to 2 or 3 in the they need to stay in the trees to be raised, as above criteria, and saved only the number of if they get near the earth vapors they fall ill entries, gazetteer name, and year. I took one and die. or more records of a given species in one gazetteer edition to equal one sighting. In a These scored gibbon data can be year with multiple publications, each edition categorized to provide time-series of with a record was counted as an individual sightings records with either greater sighting. resolution (any score) or reliability (mean and top scores). I separated it into tables of To run OLE analysis on the animal top score per year; mean score per year; and records, I used R version 3.3.0 (R Core summed score per year, to see what insights Development Team 2016) with the package into the data this might reveal when OLE is sExtinct (Clements 2013). Confidence applied. intervals can be determined during the calculations. OLE accuracy improves with 11

the number of records, requiring at least 5 and preferably more than 10 (Collen & Turvey 2009; Boakes et al 2015).

To look for a trend in how the records were distributed over time, I applied Mann- Kendall trend analysis to the variations on the gibbon data in R version 3.3.0 with the trend package (Thorsten 2016). For gazetteer publication, gibbon records scored 2, scored 3, and scored 2 or 3 and other species this is a time series of 1 for any occurrence in a given year, 0 for no occurrence. I also ran Mann-Kendall trend analysis on time series of the annual top scores, mean scores, and summed scores. 12

Results these mammal types to modern species concepts. My searches revealed 114 total references to the search terms yuán in 29 Four types of bear are listed individual gazetteers, of which 28 records (rénxiong, mǎxióng, gǒuxióng, zhūxióng) as were excluded for clearly not referring to well as the generic bear (xióng). They are gibbons, 35 were scored 1, 47 were scored 2, described as having identical bodies but and 4 were scored 3. The archaic words for different faces, probably ruling out sun bear ape xīngxing and fèifei (van Gulik 1967) are (Helarctos malayanus) which is found in the not found in the corpus. Of the 114 gibbon region. I collapsed all of these into one and references, 21 were found in sections on assumed it was Asiatic black bear (Ursus ‘beasts’ (shòu). A search of the other editions thibetanus), the species of bear known to revealed only 3 which had a ‘beasts’ section exist on Hainan today, as safe reading, given that did not mention gibbons. Four years had that e.g. the giant panda (Ailuropoda multiple publications, where gazetteers were melanoleuca) was traditionally classified as a produced on a county scale, referring to type of leopard (Li 1957). Variations on wild different parts of Hainan. pig (yězhū) and mountain pig (shānzhū) likewise are counted as referring to wild boar The initial count of other named types (Sus scrofa). Two terms (chuānshānjiǎ, of mammal in the Hainan gazetteers was 128 línglǐ) are presented as synonyms for but only 16 of these types could be pangolin (Manis pentadactyla). Both the confidently identified as corresponding to generic monkey (hóu) and more specific recognized species concepts, of which 13 are macaque (mí) are taken as mammals, and 8 both robust and plentiful (Macaca mulatta), the only monkey known enough to analyze. Most of the excluded to have occurred historically on Hainan. animals have pre-modern names and little or Some leopard accounts distinguish between no description. By way of example there are three types (bào, àibao, jīnqiánbào) which I 16 variations on lí, usually translated as have taken as (Neofelis civet, and 15 variations on shǔ, a broad term nebulosa), because this is known to exist on for rodents; these two categories together Hainan (Lau et al 2010) and all three names could cover a range of civets, mustelids, contain the generic ‘leopard’ (bào). In the small felids, rodents, and perhaps binturong, Chinese naming, as in English but not in and I could only reasonably identify 5 of zoology, all three are ‘leopards’. The 13

patterning of leopard (Panthera pardus) is Hainan.’ However, illustrated 20th century described explicitly in jīnqiánbào, which dictionaries suggest the raccoon dog is more appears 7 times, however each entry is an likely to be described as a type of lí (Hanyu identical description quoting a 16th century dacidian 1986), and the raccoon dog is not pharmacopeia (Li 1957). The lack of any known from Hainan, making it no more original information or language in the likely as a possible species identification entries suggests this is defaulting to standard based on independent evidence (Smith & Xie texts rather than reporting based on local 2008). The gazetteers record 11 types of lí to knowledge, and likely a shorthand which I could not ascribe a species, so it may interpretation for clouded leopard. be more likely that the raccoon dog, if present at all, would fit in there, leaving the The final two species that could be láng more likely to be, as it is commonly identified relatively confidently to modern understood, wolf. Where described in the mammal species concepts appear to be dhole gazetteers, a standard text from about 1,500 (Cuon alpinus) and wolf (Canis lupus) (chái, earlier is re-used: ‘like a hound, with a láng respectively). Neither is confirmed on pointed head and pale cheeks, tall at the Hainan. Supporting evidence suggests that front and wide at the rear’ (Duan 1997). the use of these names may be valid—both Dhole is recorded 13 times, from 1618 to species are or were historically present in 1935. Seven of the records share 2 standard Chinese and Vietnamese regions adjacent to descriptions, again standard texts pre-dating Hainan, and would likely have been known the gazetteer collection by more than 1,000 to the officials compiling the gazetteer years. (Smith & Xie 2008; Lau et al 2010). The 19th century naturalist Robert Swinhoe The gibbon records become more (1870) took ‘the first [dhole] to imply a frequent in later years (Fig. 2). species of Lupus, and the other [wolf] Fig. 2: Spread of gibbon records in the Hainan Nyctereutes procyonides, Gray; but it is very gazetteers, according to (a) mean score per record, (b) top score per year, (c) sum of scores per year, and (d) questionable whether either occurs in total number of mentions per year. (a) (b) 2.0 3.0 2.5 1.5 2.0 Top scores Top Mean scores 1.0 1.5 0.5 1.0

1600 1700 1800 1900 1600 1700 1800 1900

Year Year

(c) (d) 14 10 12 8 10 8 6 6 4 Individual mentions Annual sum of scores Annual 4 2 2

1600 1700 1800 1900 1600 1700 1800 1900

Year Year

This is further confirmed by Mann-Kendall trend analysis which indicates all the samples show a positive trend (Table 1) for both gazetteer publication dates and sightings not just of gibbons but of all the study-species.

Table 1: Mann-Kendall trend analysis showing Kendall’s tau, test statistic S and p-values rejecting the null hypothesis of no trend in the records over time. p-value Kendall’s tau S Gibbons: mean annual scores 0.00198 0.039 3407 Gibbons: count of annual entries 0.00348 0.035 3082 scored 2 or 3 Gazetteers publication < 2.22e-16 0.521 90832 Bear 0.00102 0.039 3372 Boar 0.000148 0.045 3896 Macaque 0.00124 0.039 3399 Pangolin 0.00336 0.036 3158 Clouded leopard 0.0146 0.033 2856 Dhole 0.0182 0.023 2022 Wolf 0.0153 0.02 1736 15

In each case the data show a positive trend, with more records in more recent times.

The OLE with 95% confidence intervals (CI) for the gibbons according to: mean score; scored 2 or 3; and summed score; and with the same CI for other species all scored equivalent to 2 and 3 (Table 2), shows the predicted survival of all species past the final gazetteer record, i.e. 1935.

Table 2: The Optimal Linear Estimation (OLE) of the year a species went extinct, with 95 % CI, and all species other than gibbons scored equivalent to 2 and 3. Species Estimated Lower CI Upper CI Number of extinction sightings Gibbons, by mean annual score: 1938 1935 1961 23 Gibbons, scored 3* 1898 1829 5260 4 Gibbons, scored 2 1938 1935 1961 16 Gibbons, scored 2 or 3 1938 1935 1957 20 Gibbons, sum of scores 1938 1935 1958 23 Gibbons, scored 1, 2 or 3 1937 1935 1950 26 Bear 1938 1935 1961 20

Boar 1936 1935 1948 23 Pangolin 1938 1935 1959 22 Macaque 1937 1935 1956 22 Clouded leopard 1938 1935 1959 27 Dhole 1942 1935 1999 14 Wolf 1953 1932 2062 9

* Only 4 records were scored 3, making it inadequate for reliable OLE, which requires at least 5 entries. It is included here just for completeness. Discussion (Solow 2005) based on a Weibull distribution, a two-parameter model from This study suggests the Hainan weakest-link analysis (Crawley 2005). gazetteer can be used as a robust source Accuracy improves with the number of towards developing historical baselines of records, requiring at least 5 and preferably the mammals found on Hainan. Thus more than 10 (Collen & Turvey 2009; validated, the gazetteers also suggest the Boakes et al 2015). In this dataset there were presence of two previously unrecorded large 387 years between 1521 and 1935 with no carnivores, wolf and dhole, into the 20th publications, which could be seen as times century. This may have implications on how when the observation effort was zero. Larger the ecosystem of the island is understood, gaps between sightings lead to more which in turn may be relevant to Hainan conservative extinction estimates, however gibbon conservation action. the method likewise does not assume that observation effort is constant, just remaining A statistical approach to validating above zero (Collen et al 2010; Boakes et al such historical sightings often has to work 2015). Boakes et al (2015) say this is largely with very limited data, given, in the case of unimportant provided the sightings are Hainan gibbons, no reliable population comparatively rare. Given the information studies for decades after the last gazetteer available, OLE appears to be an appropriate record (Turvey at al 2015a). A reference tool to assess the gazetteers. point may come from non-statistical data. Observations today can help infer historical The OLE analysis indicates presence: if a species is not extinct on an persistence for species beyond the end of the island today, it was likely not yet extinct in gazetteer record, with extinction estimates of 1935. I compared the gazetteer dataset with 1936 to 1953, and upper CIs of at least 12 tabulated strengths and weaknesses of years after the last sighting. This includes the models to estimate extinction (Boakes et al still extant gibbon, bear, and pangolin (Lau et 2015). OLE is a useful tool to assess al 2010; Turvey 2015a; Nash et al 2016), extinctions, be they global or local, from just suggesting the gazetteers are a reliable source time-series sightings data (Collen et al 2010; within their own timeframe. Species that Boakes et al 2015). It is a probabilistic appear in the records but are known not to method, which estimates a year of extinction have existed on historical Hainan, such as as a weighted-sum of the sighting times tiger (hǔ), rhino (sì), and water monster (yù) 17

(Schafer 1970, Lau et al 2010) are filtered influencing the officials who wrote the out by the qualitative scoring process, thus gazetteers: political and educational norms species that I put through OLE are all more change and may affect the scientific plausibly present at last sighting than not. awareness of the authors, or their likelihood This adds validation to the findings of of quoting older sources, and thus influence previous studies which have been based on the quality of individual records. Such a the gazetteer record (Chatterjee et al 2012; system may further overcome the effects of Turvey et al 2015b). higher record densities in later years, and would contribute to the GIS and statistics- The formulaic nature of the gazetteer based Chinese primate extinction early- as a genre provides little scope for agency on warning system called for by Pan et al (2016) the part of the individual observer (Pooley by increasing the resolution of historical 2013): the official who writes the account baselines. However such an inter-disciplinary completes the same ‘produce’ section of the process would be highly resource-intensive, ‘geography’ chapter referencing the same and studies made in China may be subject to standard works as his predecessors did. It contemporary political concerns. might be hoped that widening the search to include other chapters on literature, Improved spatial resolution may also biography, and trade, would escape this and be possible. The gazetteers are published for perhaps include fresh detail. In the Hainan administrative units ranging from county, to gazetteers, this is only partially the case, island-wide, to province. My attempts to because while they did enlarge the dataset, identify the historical boundaries met none of my records from outside the contradictory reports between different geography or territory chapters scored above gazetteers and historical mapping services 2. (CHGIS 2007). However there can be consistency within a given dynastic reign. The Mann-Kendall trend analysis This may provide some potential for further shows more records in more recent years. I defining geographical spread, but over sought to overcome this by the qualitative shorter temporal scales. method of scoring entries for reliability. Further qualitative approaches may bring The greatest challenge to using the greater data resolution. It would be possible gazetteers to establish historical baselines is to temporally map the broad trends identifying species. In this study that 18

excludes 50 types of animal recorded but The more questionable is wolf, first recorded difficult to identify, such as the 16 rodent- in 1707, continuing until 1931. However like (shǔ). The deer of Hainan are a only a standard description ever appears, and illustrative example: The gazetteers indicate this is from a dictionary complied in 121 CE 12 names for types of deer, plus variations (Duan 1997; Wilkinson 2015). No mention is for sex. Accounts will claim two terms as made of the typical wolf behavior of hunting being alternative names for the same type, in packs (Macdonald 2009), while some of but not consistently. Some clarity comes in the deer references do describe large herds, the form of western scientific studies from so this behavior is among the characteristics the 19th and 20th centuries (Swinhoe 1870; that the authors could have noted. A 1774 Ohtaishi & Gao 1990), but even so this entry in the gazetteers specifically states that requires more work with additional sources there are no wolves on Hainan—but the very before drawing taxonomic lines. same edition also contains a record of wolf presence, as do 8 other editions. This The apparent robustness of the presents an opportunity for further qualitative gazetteers for identifiable species has research looking at the overall and zoological implications. Wolf and dhole are recorded reliability of these 9 editions, an exercise and apparently identifiable, despite neither which could further test OLE as a method of being scientifically confirmed on Hainan assessing gazetteer robustness. Including (Swinhoe 1870; Lau et al 2010). On a global Hainan in the historical wolf distribution scale, that a large mammal carnivore should maps would mean extending the range south be extirpated from an area in the 20th century by about 800 km (Mech & Boitani 2010). is on-trend: large mammals are twice as likely to be threatened as smaller, while in Dhole is recorded 14 times, across China the range of tigers has shrunk 95% almost the entire time-span of the gazetteer (Macdonald et al 2013). With its large scale collection. More than half the records share 2 land-use change (Zhang et al 2010), Hainan descriptions: 5 say ‘Like a dog but with a can conceivably follow the same pattern. longer tail’ while 2 go into more detail: ‘A type of wolf, deep fur and dog-like feet, a The gazetteers record wolf and dhole long tail, pale cheeks, colored 9 and 14 times respectively. The OLE results yellow/brown’. Published in 1875 and 1931 suggest that wolf and dhole may have these come after Swinhoe (1870) dismissed persisted on Hainan into the 20th century. the likelihood of dhole on the island. The 19

historical range maps of dhole just miss out encountered cryptic species (Murray 2011). Hainan, reaching the nearby landmass of The gazetteers may again inform ecology Vietnam (Kamler et al 2015), and the research because as well as the mammals in presence of dhole on Hainan into the 20th this study, they also list plants, trees, century seems likely. and fish over the centuries, and at least one account of the wild animal pelts found at The ecosystem may provide clues as markets run by the indigenous Lí people. to whether dhole, wolf, or both were on Hunting by humans is reported to exert an Hainan within the past 100 years. Both influence in ecosystem cascades involving would have had prey in the form of the large carnivores such as wolves and bears island’s deer and small mammals (Hayward (Ripple et al 2014), and ungulates. et al 2014). This addition of two large carnivores to the confirmed bears and Taken directly, the Hainan gazetteers clouded leopard would likely have provide little more than sightings data. implications for the ecosystems that were in However, this study shows that they can be a place on Hainan in the early 20th century robust record of the historical presence of (Ripple et al 2014). There may be need for a animals, and can be used to inform radical re-think of habitat restoration targets conservation management planning. At least on Hainan, including possible trophic knock- 2 of the secondary study species here, wolf on effects for Hainan gibbon food trees, due and dhole, are not reported on Hainan today. to carnivores controlling the herbivore The main study species, the Hainan gibbon, population and dispersal (Marris 2014; is the rarest ape in the world. With fewer Ripple 2014; Turvey et al 2015a). There are than 30 remaining it is an extreme case but nearly as many written sources of Chinese with anthropogenic pressures, including history from 1911 to 1949 as there are in the climate change, increasing, it is unlikely to previous 3 millennia, including archives and be the last example of a species facing newspapers, in both the People’s Republic of extinction in Hainan or China. As such China and Taiwan (Wilkinson 2015). including the gazetteer record to help Possible 20th century sources to examine establish a historical baseline is a meaningful include local newspapers and the accounts of process. In the case of Hainan it is valid, and communist fighters, who were based in the throughout China both likely to be needed mountains before 1949 and may have and worth repeating for other species. Supporting Information description

Chinese characters and transliteration (Appendix S1), Digitization workflow (Appendix S2) Raw data: gibbons (Appendix S3), and Raw data: other species (Appendix S4), are available online. The authors are solely responsible for the content and functionality of these materials. Queries (other than absence of the material) should be directed to the corresponding author. 21

Literature cited

Anthony L. 2014. AntConc. Waseda University, Tokyo, Japan. Boakes EH, Rout TM, and Collen B. 2015. Inferring species extinction: the use of sighting records. Methods in Ecology and Evolution 6(6):678-687. Bol PK. 2001. The Rise of Local History: History, Geography, and Culture in Southern Song and Yuan Wuzhou. Harvard Journal of Asiatic Studies 61(1):37-76. Chatterjee HJ, Tse JSY, and Turvey ST. 2012. Using Ecological Niche Modelling to Predict Spatial and Temporal Distribution Patterns in Chinese Gibbons: Lessons from the Present and the Past. Folia Primatologica 83(2):85-99. CHGIS January 2007. China Historical GIS. Harvard Yenching Institute, Cambridge, MA. Available from http://www.fas.harvard.edu/~chgis/data/chgis/downloads/v4/ 2016. Cihai: hedingben. 1948. 2nd edition. Zhonghua shuju, Shanghai. Clements C. 2013. sExtinct. 1.1st edition. Collen B, and Turvey ST. 2009. Probabilistic methods for determining extinction chronologies. Pages 181- 191 in ST Turvey, editor. Holocene Extinctions. Oxford University Press, Oxford. Collen B, Purvis A, and Mace GM. 2010. BIODIVERSITY RESEARCH: When is a species really extinct? Testing extinction inference from a sighting record to inform conservation assessment. Diversity and Distributions 16(5):755-764. Crawley MJ. 2005. Statistics: an introduction using R. John Wiley, Chichester. Davies AL, Colombo S, and Hanley N. 2014. Improving the application of long-term ecology in conservation and land management. Journal of Applied Ecology 51(1):63-70. Duan Y. 1997. zhu=Annotated Shuowen jiezi. 1st edition. Jiangsu guangling gujikeyinshe, Yangzhou. Edmonds RL. 1999. The Environment in the People's Republic of China 50 Years On. The China Quarterly 159:640-649. Fleck DW, Voss RS, and Patton JL. 1999. Biological Basis of Saki (Pithecia) Folk Species Recognized by the Matses Indians of Amazonian Perú. International Journal of Primatology 20(6):1005-1028. Guhanyu changyongzi zidian bianxiezu. 1998. Guhanyu changyongzi zidian: 1998 nian ban. Shangwu yinshuguan, Beijing. Gulik RHv, 1910-1967. 1967. The gibbon in China: an essay in Chinese animal lore. Guo P. 1981. yintu. 1st edition. Guangwen shuju, Taibei. Hanyu dacidian. 1986. Di 1 ban. edition. Shanghai cishu chubanshe, Shanghai. Hayward MW, Lyngdoh S, and Habib B. 2014. Diet and prey preferences of dholes (Cuon alpinus): dietary competition within Asia's apex predator guild. Journal of zoology 294(4):255-266. Hong S., and Hainan difang wenxian congshu bianweihui. 2003. Hainan difangzhi congkan = Collected Hainan Gazetteers. Hainan chubanshe, Haikou. Kamler JF, Songsasen N, Jenks K, Srivathsa A, and Sheng, L. & Kunkel, K. 2015. Cuon alpinus. The IUCN Red List of Threatened Species 2015 (e.T5953A72477893):15 August 2016. Lau MW, Fellowes JR, and Chan BPL. 2010. Carnivores (Mammalia: Carnivora) in South China: a status review with notes on the commercial trade. Mammal Review 40(4):247-292. Li S. 1957. Bencao gangmu. 1st edition. Renmin weisheng chubanshe, Beijing. 22

Liu J, and Diamond J. 2005. China's environment in a globalizing world. Nature 435(7046):1179-1186. Macdonald DW, Boitani L, Dinerstein E, Fritz H, and Wrangham R. 2013. Conserving large mammals: are they a special case? Pages 276-312 in D. W. Macdonald and K. J. Willis, editors. Key Topics in Conservation Biology 2. John Wiley and Sons Ltd, Chicester. Macdonald DW. 2009. The encyclopedia of mammals. New edition. Oxford University Press, Oxford. Marris E. 2014. Rethinking predators: Legend of the wolf. Nature 507(7491):158-60. Mech LD, and Boitani L. 2010. Canis lupus. The IUCN Red List of Threatened Species 2010 2010(e.T3746A10049204):Downloaded on 15 August 2016. http://dx.doi.org/10.2305/IUCN.UK.2010- 4.RLTS.T3746A10049204.en.1. Murray JA. 2011. Culturing revolution: the local Communists of China's Hainan Island. UC San Diego. Nash HC, Wong MHG, and Turvey ST. 2016. Using local ecological knowledge to determine status and threats of the Critically Endangered Chinese pangolin (Manis pentadactyla) in Hainan, China. Biological Conservation 196:189-195. Ohtaishi N, and Gao Y. 1990. A review of the distribution of all species of deer (Tragulidae, Moschidae and Cervidae) in China. Mammal Review 20(2-3):125-144. Pan R, Oxnard C, Grueter CC, Li B, Qi X, He G, Guo S, and Garber PA. 2016. A new conservation strategy for China—A model starting with primates. American Journal of Primatology. Pauly D. 1995. Anecdotes and the shifting baseline syndrome of fisheries. Trends in Ecology & Evolution 10(10):430 http://dx.doi.org/10.1016/S0169-5347(00)89171-51. Pooley S. 2013. Historians are from Venus, Ecologists are from Mars. Conservation Biology 27(6):1481- 1483. R Core Team. 2016. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. Rick TC, and Lockwood R. 2013. Integrating Paleobiology, Archeology, and History to Inform Biological Conservation. Conservation Biology 27(1):45-54. Ripple WJ, Estes JA, Beschta RL, Wilmers CC, Ritchie EG, Hebblewhite M, et al. 2014. Status and Ecological Effects of the World’s Largest Carnivores. Science 343(6167): 10.1126/science.12414841. Roberts DL, and Solow AR. 2003. Flightless birds: When did the dodo become extinct? Nature 426(6964):245-245. Schafer EH. 1970. Shore of pearls. Univ. of California Press, Berkeley and Los Angeles. Schipper J, Chanson JS, Chiozza F, Cox NA, Hoffmann M, Katariya V, et al. 2008. The Status of the World's Land and Marine Mammals: Diversity, Threat, and Knowledge. Science 322(5899):225-230 10.1126/science.11651151. Smith AT, and Xie Y, editors. 2008. A guide to the mammals of China. Princeton University Press, Princeton, N.J. Solow AR. 2005. Inferring extinction from a sighting record. Mathematical Biosciences (195):47-55. Swinhoe R. 1870. The natural history of Hainan. Horace Cox, London. Szabó P. 2010. Why history matters in ecology: an interdisciplinary perspective. Environmental Conservation 37(04):380-387. Thorsten P. 2016. trend: Non-Parametric Trend Tests and Change-Point Detection. 23

Turvey, ST, Traylor-Holzer, K, Wong, MHG., Bryant, JV, Zeng, XY, Hong, XJ, Long, YC. 2015a. International Conservation Planning Workshop for the Hainan Gibbon: Final Report. Zoological Society of London/IUCN SSC Conservation Breeding Specialist Group, London & Apple Valley, MN. Turvey S. 2009. Witness to extinction: how we failed to save the Yangtze River dolphin. Oxford University Press, Oxford. Turvey ST, Crees JJ, and Di Fonzo MMI. 2015b. Historical data as a baseline for conservation: reconstructing long-term faunal extinction dynamics in Late Imperial–modern China. Proceedings of the Royal Society of London B: Biological Sciences 282(1813): 10.1098/rspb.2015.12991. Wen R. 2013. Zhongguo gudai yeshengdongwu dili fenbu = Geographical distribution of wild animals in ancient China. Wen R. 2009. Zhongguo zhenxi yeshengdongwu fenbubianqian = Changes in distribution of China’s endangered wild animals. Wilkinson EP. 2015. Chinese history: a new manual. Harvard University Asia Center, Harvard University Press, Cambridge, Mass. Zhang M, Fellowes JR, Jiang X, Wang W, Chan BPL, Ren G, and Zhu J. 2010. Degradation of tropical forest in Hainan, China, 1991–2008: Conservation implications for Hainan Gibbon (Nomascus hainanus). Biological Conservation 143(6):1397-1404. 24

Appendix S1. Chinese characters and transliteration

This paper shows simplified Chinese characters, as these were used in the received version of the base text. Transliterations into the Roman alphabet are in Hanyu Pinyin. Both these are also the ISO standards. Glossary: Hainan 海南 Hǎinán Yangtze river dolphin or baiji 白暨 báijì Local gazetteer 地方志 dìfāngzhì Ape, gibbon 猿 yuán and 猨 yuán Ape (archaic) 猩猩 xīngxing; 狒狒 fèifei Beast, approximating mammal: 兽 shòu Civets etc. 狸 lí Rodents etc.鼠 shǔ Bears:人熊 rénxiong; 马熊 mǎxióng; 狗熊 gǒuxióng; 猪熊 zhūxióng; generic 熊 xióng Boar 山猪 shānzhū，野猪 yězhū Pangolin 穿山甲 chuānshānjiǎ, 鲮鲤 línglǐ Monkey 猴 hóu Macaque: 猕 mí Leopard: 豹 bào Clouded leopard: 金钱豹; jīnqiánbào 艾豹 àibào Dhole: 豺 chái Wolf 狼 láng Geography 地理志 dìlǐzhì Territory 舆地志 yúdìzhì Produce 物产 wùchǎn Yu water monster 蜮 yù Li (ethnic group) 黎 Lí Tiger 虎 hǔ Rhino 兕 sì 25

Quoted passages: In addition, Qiongzhou [Hainan] has stone apes, the small ones the size of a fist, they grow when fed water from a well. Also [known as] black apes, they can rub ink [from an inkstone], jumping into the ink pot when finished. 琼州又有石猨，小者拳许，饮以井水即长。又墨猨能磨墨，磨 毕跳入笔筒中。

Gibbons: there are three types: Golden ones are yellow, jade-faced are black, and there are also pure black ones. The golden and jade-faced ones are all rare. 猿 有三种。金丝者黄，玉面者黑，又有 纯黑者。金丝、玉面皆难得。

Gibbons: the male is black and the female pale. They are like monkeys but their arms are very long, climbing and grabbing branches in the canopy with great agility. If they ever fall to the ground they stiffen up like a tree. As they are good at climbing trees but cannot walk, they need to stay in the trees to be raised, as if they get near the earth vapors they fall ill and die. 猿雄黑雌白，似猴而两 臂甚长，攀树援岭往来甚捷，一堕平地，则木强矣。以善援木而不能走也，畜之亦须置树

间，近土气即病而死

Like a dog but with a longer tail 豺似狗而长尾 A type of wolf, deep fur and dog-like feet, a long tail, pale cheeks, colored yellow/brown 豺狼属， 深毛而狗足，长尾，白颊，色黄

Chinese literature cited:

洪寿祥 Hong Shouxiang. 2003. 海南地方志丛刊 Hainan difangzhi congkan. 海南地方志丛刊 Hainan difangzhi congkan. 1st ed.

Haikou: 海南出版社 Hainan Chubanshe. 辞海 Cihai: Hedingben.1948. ed. Zhonghua shuju. 2nd ed. ed. Shanghai: Zhonghua shuju. 古汉语常用字字典编写组 Guhanyu changyongzi zidian bianxiezu., 1998. 古汉语常用字字典 Guhanyu changyongzi zidian : 1998 nianban. 北京 Beijing: 商务印书馆 Shangwu yinshuguan. 汉语大词典 Hanyu dacidian 1986. eds. 汉语大词典编辑委员会；主编罗竹风 Zhufeng Luo, Hanyu dacidian bianji weiyuanhui. and 汉语大词典编纂处 Hanyu dacidian bianzuanchu. Di 1 ban. ed. 上海 Shanghai: 上海辞书出版社 Shanghai cishu chubanshe.

文榕生. 2009. 中国珍稀野生动物分布变迁. 山东科学技术出版社, Shandong.

26

Appendix S2. Digitization workflow

In 2003 the Hainan Publishing house compiled all extant Hainan gazetteers, editing them into a published collection. However they did so only in a 68 volume printed edition, not making it available digitally. An unofficial ebook version of many Chinese historical gazetteers is available from an antiquarian book seller on China’s leading shopping website, Taobao.com. These are delivered on demand as DVDs containing scans of the paper book, supplied as individual images of each page, organized in folders equivalent to each volume. I had to make these searchable by running the 1.9 GB of image data through optical character recognition software FineReader Pro (ABBYY 2016): at this stage I found 23 volumes were corrupted and 6 missing, and discarded them as bad data. I found one missing volume online as a lower-resolution PDF, so it was included in the corpus. In this volume the OCR error rate is likely to be higher than for the others, which have higher image-quality. This left 39 volumes to work with. The western or Gregorian calendar was introduced to China after publications in this dataset. I converted each volume’s traditional dates to calendar years, based on the year of printing or the most recent revision, as listed in the Forward of each gazetteer. The Forwards may variously state the year the book was commissioned, revised, or printed. There is no way of telling for sure when between those years the animal sightings were actually made–or, indeed, whether they were seen at all. 27

Appendix S3. Raw data: gibbons

The spreadsheet recording gibbon mentions had the following column headings: Id; Gazetteer name; hanzi county; county; hanzi date; date; Species accounts; Specific region; Area today; Original text (yuan); Translation; Referring to; txt file; PNG; Other species listed in adjacent text (or animal list if no yuan in volume); Rank; Other species text; yuanYN. In the attached file some columns have been removed to accommodate the single PDF file submission requirement. Complete Excel spreadsheets are available from the author on request.

28

Appendix S4. Raw data: other species

These were collected in the paragraphs surrounding gibbon mentions, and in the ‘beasts’ sections. In rare cases these sections include non-mammal species.